Fan-Yun Sun (@sunfanyun) / X

Fan-Yun Sun

355 posts

Fan-Yun Sun

@sunfanyun

ceo @moonlake, prev. @nvidiaai, cs phd @stanfordailab

Stanford, CA

ai.stanford.edu/~sunfanyun/

Joined October 2018

Pinned
Fan-Yun Sun
@sunfanyun
Jun 3
Excited to share that we are working with Nvidia to bring @moonlake's 3D Agent capabilities into NVIDIA's simulation ecosystem (@nvidiaomniverse) to help teams build assets, digital twins, and scenarios at scale. Here are 8 pairs of input images and the generated
00:00
916K
Fan-Yun Sun
@sunfanyun
Oct 12, 2025
Works like this deserve a lot more attention. In embodied AI, the most common argument against physics-based simulation is that deformables are hard/expensive to simulate. Offset Geometric Contact just made it 300x faster. It computes vertex-specific displacement bounds to
00:00
156K
Fan-Yun Sun
@sunfanyun
Nov 24, 2024
Congratulations to the Edify team @NVIDIAAI on the release of their text/image-to-3D model Besides the enhanced visual fidelity, I think the most impressive thing is how clean the quad mesh is compared to other state-of-the-art text/image to 3D models! This has significant
00:00
68K
Fan-Yun Sun
@sunfanyun
Oct 1, 2025
I told my parents that I’d like to drop out of my cs phd program at Stanford a few months back. They didn’t let me, we’re asian :) So I graduated and started @moonlake. @sharonal_lee and I saw urgency, and opportunities. Excited to share that we raised a 28 million dollar
01:00
Moonlake
@moonlake
Oct 1, 2025
We raised $28M seed from Threshold Ventures, AIX Ventures, and NVentures (Nvidia's venture capital arm) —alongside 10+ unicorn founders and top AI researchers— to build reasoning models that generate real-time simulations and games. Models are bottlenecked by practical
156K
Fan-Yun Sun
@sunfanyun
Aug 4, 2024
This is probably my favorite definition of intelligence (this figure is taken from @fchollet's paper "On the Measure of Intelligence"). "Higher intelligence covers more ground in future situation space using the same information." LLMs are not intelligent, but they are still
38K
Fan-Yun Sun
@sunfanyun
Dec 18, 2023
Our research introduces a system that enables you to generate 3D environments from text prompts and train embodied AI agents within them! Website: yueyang1996.github.io/holodeck/ Code: github.com/allenai/Holode… How did we leverage Objaverse assets to create interactive 3D environments? 👇
00:00
36K
Fan-Yun Sun
@sunfanyun
Mar 18, 2025
Spatial reasoning is a major challenge for the foundation models today, even in simple tasks like arranging objects in 3D space. #CVPR2025 Introducing LayoutVLM, a differentiable optimization framework that uses VLM to spatially reason about diverse scene layouts from unlabeled
00:00
93K
Fan-Yun Sun
@sunfanyun
Sep 28, 2024
Training RL/robot policies requires extensive experience in the target environment, which is often difficult to obtain. How can we “distill” embodied policies from foundational models? Introducing FactorSim! #NeurIPS2024 We show that by generating prompt-aligned simulations and
00:00
35K
Fan-Yun Sun
@sunfanyun
Feb 26, 2025
Claude 3.7 one-shotted this 3D room— everything from layout to geometry and texture. This is always a fun way to test a model’s spatial reasoning ability
00:00
33K
Fan-Yun Sun
@sunfanyun
Oct 6, 2025
😅 we are hiring, DM open
Hsuan Su
@jacksukk
Oct 6, 2025
Came to a Stanford event with @sunfanyun.. couldn’t get him to leave because too many recruits were trying to slide… @moonlake
26K
Fan-Yun Sun
@sunfanyun
Oct 23, 2025
If you work on real-time models, code generation, RL, or ML infra at Meta and are affected by re-org, apply to work with us We already have something better than Horizon Worlds
Moonlake
@moonlake
Oct 1, 2025
Replying to @moonlake
Moonlake’s models reason through the representation, tools, and models to use to generate interactive experiences. Let us walk you through a generation:
00:00
32K
Fan-Yun Sun
@sunfanyun
Oct 16, 2025
Trillions will be poured into training action-conditioned world models in the coming decade. The bottleneck is in *action-conditioned* data, and video game is the where this type of data naturally accrue today. GENIE is already trained on the largest video game dataset that
4.6K
Fan-Yun Sun
@sunfanyun
Nov 24, 2024
Replying to @sunfanyun
Their demo website here: research.nvidia.com/labs/dir/edify…
00:00
4K
Fan-Yun Sun
@sunfanyun
Oct 12, 2025
Replying to @sunfanyun
paper here: graphics.cs.utah.edu/research/proje…
5.3K